The increased popularity of personal mobile communication and of portable computing will significantly increase the demand for wireless communication capacity, since they are heavy users of e-mail and web-browsing applications. Access through fixed wireless communication will also contribute to this demand growth. Voice calls will increase the circuit-switched traffic loads that have dominated wireless communication up until now. The emerging convergence of voice and data traffic into a single flow of IP packets will generate an ever-increasing packet-switched traffic load. Resource utilization efficiency is key to preserving the dynamic multiplexing advantages achievable from this convergence. Regardless of its origin, the increased demand will require more efficient methods of utilizing the RF spectrum allocated for this purpose.
Digital multiple access techniques must be optimized in order to meet the growing demand for wireless communication capacity. The concern here is with digital “channelized” methods, such as the North American time division multiple access (TDMA)/frequency division multiple access (FDMA) system defined by the IS-136 digital cellular standard or the European TDMA/FDMA system defined by the Groupe Speciale Mobile (GSM) digital cellular standard. A channel in a TDMA/FDMA system is a time-slot. The objective in maximizing traffic-carrying capacity is to assign a channel to the unit of traffic load in a way that maximizes the system's throughput without violating quality-of-service (QoS) requirements. This is true whether it is a circuit-switched call that has exclusive use of the circuit or a packet-switched call that shares the circuit with other calls.
New algorithms must be devised the meet an array of different quality-of-service (QoS) requirements that arise with packet-switched traffic from diverse applications. Though the QoS requirements can differ between different traffic types, their common objective and environmental/engineering constraints permit the use of similar channel assignment concepts for both. This is because the assignment of channels to a packet in a packet-switched network resembles the assignment of a channel to a call for its duration in a circuit-switched network. To underscore this similarity, the terms packet and call are used interchangeably herein.
The notion of applying circuit-switched channel assignment concepts to packet-switched traffic should be pursued with care as there exist fundamental differences. An important consideration relates to the asymmetry of the assignment in the two communication directions, the uplink direction from the mobile to the base station and the downlink direction from the base station to the mobile. With circuit-switched voice traffic, the uplink and downlink channels are paired, and channel assignment can thus utilize information available on only one direction. With packet-switched traffic, uplink and downlink channels would be typically independently assigned. A reason is the imbalance of the traffic loads in the two directions: there is more traffic on the downlink than on the uplink for certain applications, such as web browsing. But, even if the loads were balanced, channel assignment should occur independently in the two directions because of the non-coincidence of uplink and downlink packets, even when occurring during the same session.
Uncoupling the channel assignment for the two communication directions affects the applicability of circuit-switched traffic algorithms to packet-switched traffic. An important consideration relates to channel assignment algorithms relying on interference-sensing, known also as measurement-based algorithms. The problem to be addressed is how to perform asymmetric measurement-based channel assignment on a single direction.
Another consideration in packet-switched channel assignment is the packet length. Short “calls” require that the time dedicated to channel assignment be short in order to avoid capacity loss. To carry out channel assignment efficiently in a multi-cell environment, the multiple base stations must be synchronized. With packets of constant length, or of a small-integer multiple of a fixed length, channel selection by different base stations will occur simultaneously. In that case the possibility of contention for the same channel becomes likely.
I. Measurement Based Dynamic Channel Assignment (DCA)
There are two general forms of flexible channel assignment: adaptive, dynamic, and their combination known as adaptive-dynamic. See for example “Channel Assignment Schemes for Cellular Mobile Telecommunication Systems: A Comprehensive Survey”, I. Katzela and M. Naghshineh, IEEE Personal Communications, June 1996. Also see the above referenced U.S. Pat. No. 5,404,574 and U.S. Pat. No. 5,809,423. Adaptive channel assignment (ACA) is a time-variable fixed channel assignment (FCA), whereby the set of channels allocated for use by a base station is fixed over a time interval, typically longer than the call duration. Dynamic channel assignment (DCA) also involves allocated sets of channels, whose membership may vary in time. The difference between the two is that the channel sets in ACA do not overlap and, more importantly, all allocated channels can be used simultaneously without violating the pre-specified quality-of-service target. With DCA, on the other hand, this requirement on the composition of channel sets is not observed. In fact, a common DCA approach employs channel sets comprising all the available channels. Consequently, before a channel is assigned to a call by DCA, a test is required to establish observance of the QoS requirement. The channel-assignment admissibility (channel assignment admissibility (CAA) criteria employed in this test vary.
The channel assignment admissibility (CAA) criterion may rely either on real-time channel utilization information shared across neighboring base stations and combined with knowledge of the interference relationships between base stations, or on real-time interference sensing. Dynamic channel assignment (DCA) algorithms employing the latter criterion are referred to as measurement-based algorithms. Such algorithms are particularly attractive for distributed system architectures as they require no real-time information sharing between base stations on channel utilization.
A. Measurement-Based Channel Assignment Admissibility (CAA) Criterion
A measurement-based channel assignment admissibility (CAA) criterion delivers the target quality of service only if the utilized measurements can predict reliably whether the channel assignment under consideration will violate the quality-of-service requirement. Many of the proposed measurement-based dynamic channel assignment (DCA) algorithms fail to do so. The IS 136 digital cellular system uses the mobile stations to measure the signals from surrounding base stations and report those measurements back to the serving base station. This is primarily used for mobile assisted handoff so that the network can decide whether a handoff is required. For channel assignment in an IS 136 digital cellular system, the quality-of-service requirement will not be met when the serving base station selects a channel for an incoming voice call by relying exclusively on measurements of the signal strength on downlink channels. Such measurements for channel assignment are called Mobile-Assisted Channel Assignment (MACA) measurements.
This situation is shown in FIG. 1, which depicts an indoor cellular system and an outdoor cellular system, respectively. In FIG. 1, mobile station 101 (MS1) is registered on base station 102 (BS1) and mobile station 103 (MS2) is registered on base station 104 (BS2). In FIG. 2, mobile station 201 (MS1) is registered on base station 202 (BS1) and mobile station 203 (MS2) is registered on base station 204 (BS2). The indoor system employs omni-directional base stations (with full-aperture transmitters and receivers) housed within a building whose walls, or other obstructions, may attenuate the transmitted signal. The outdoor system employs base stations with directional transmitters and receivers that limit the coverage angle. In both systems, mobiles have omni-directional receivers and transmitters. The common feature in the two systems is that a downlink signal from BS2 102 or 202 is hardly perceptible by mobile MS1 101 or 201 because of the location of the two. An obstruction attenuates the signal in the indoor system, and the mobile lies outside the beamwidth of the directional antenna of base station BS2 in the outdoor system.
Suppose that Mobile MS2 103 or 203 is engaged in a call. Suppose also that a call is initiated by mobile MS1 101 or 201 and that the list of channels sent to mobile MS1 101 or 201 by base station B1 102 or 202 for MACA measurement includes the channel used by mobile MS2. Because of its location, the signal measured by mobile MS1 101 or 201 on the channel used by mobile MS2 103 or 203 will be sufficiently weak to lead to the assignment of that channel to the incoming call. This assignment will cause interference to mobile MS2 103 or 203, and hence violate the quality-of-service requirement, because the CM criterion used was inadequate. In general, a CM criterion that relies on mobile measurements alone to sense interference will be inadequate.
A measurement-based channel assignment admissibility (CM) criterion would meet the QoS requirement if measurement of the interference potential of a channel assignment is made on the exact same path to be traversed by the signals resulting from the assignment. In general, two types of measurements are needed in order to clear a channel for assignment. A measurement clearing the path between the mobile and the neighboring base stations, and another between the serving base station and the mobiles served by neighboring base stations. Clearance of channels can be accomplished differently depending on whether uplink and downlink channels are paired or not.
II. Coupled Uplink and Downlink Channel Assignment
Channel assignment in circuit-switched wireless networks occurs in pre-defined pairs of uplink and downlink channels. Such circuit-switched wireless networks are typically used to carry voice traffic. In such systems a measurement-based channel assignment admissibility (CAA) criterion could rely on a single signal strength measurement for both channels of the pair. See for example “Distributed Packet Dynamic Resource Allocation (DRA) for Wireless Networks”, J. F. Whitehead, Proc. of VTC '96, pp 111–115. That is, both the base station and the mobile are engaged in measurement of the signal strength of a candidate pair of channels. The mobile's measurement clears the downlink channel and its associated uplink channel along all paths between the mobile and the neighboring base stations. The base station's measurement clears the uplink channel and its associated downlink channel along the path between the serving base station and the mobiles served by neighboring base stations.
It can be seen in FIGS. 1 and 2 that the coupled measurements would provide the indication of the interference potential for the channel pair. As explained earlier, a signal strength measurement by mobile MS1 101 or 201 alone indicates that there is no call on the downlink channel used by mobile MS2 103 or 203 because the obstruction attenuates the signal in the indoor system, and the mobile lies outside the beam-width of the directional antenna of base station BS2. If, in addition to mobile MS1 101 or 201 measuring the signal strength on the downlink channel used in mobile MS2's 103 or 203 call, base station BS1 102 or 202 also measured the signal strength on the paired uplink channel, there would be the indication that the pair of channels was used by a neighboring base station and hence the assignment would not be made.
III. Directionally-Uncoupled Channel Assignment
Unlike in circuit-switched wireless systems, channel assignment in packet-switched wireless systems requires uncoupling along the uplink and downlink communication directions, as the traffic in the two directions is non-coincident. That is, packets from the mobile and the base station occur at different times. Thus uncoupling would be required even if the traffic loads in the two directions were reasonably balanced.
In theory, both frequency-division and time-division duplex can support uncoupled channel assignment between uplink and downlink. Time-division duplex allows more efficient channel training, and allocation of the radio resource between the two directions requires no planning. With frequency-division duplex, a different number of channels would be made available in the two directions when the traffic loads along the two directions differ. Additionally, the duration of the measurement and channel selection must be short for, if it is significant relative to the duration of a call, it would cause capacity loss.
The problem of accommodating unbalanced packet traffic loads by applying a measurement-based channel assignment algorithm on a single direction has been addressed in “An OFDM-Based High-Speed-Data (HSD) Air Interface Proposal”, J. C. Juang and S. Timuri, AWS submission to the Universal Wireless Communications Consortium UWCC GTF HSD/97.10.0709, Nov. 11, 1997. See also “Dynamic Packet Assignment for Advanced Internet Cellular Services”, J. C. Chuang and N. R. Sollenberger, Proc. Of Globecom '97. See also “Advanced Cellular Internet Service (ACIS)”, L. J. Cimini, Jr., J. C. Chuang, and N. R. Sollenberger, IEEE Communications Magazine, October 1998. A time division multiple access (TDMA)/frequency division multiple access (FDMA) frame structure was proposed for downlink packet assignment, which was based on the Orthogonal Frequency Division Multiplexing (OFDM) technique for multi-carrier modulation. Pilot tones that correspond to the downlink traffic channels in use are transmitted simultaneously by the base stations, thus enabling the mobiles to scan the pilots and complete channel assignment quickly. Mobiles with pending packets measure the tones transmitted by the neighboring base stations and report the list of interference-free channels to their serving base stations. The serving base station then notifies the mobiles of the channel assigned to transmit the traffic packet. By staggering the time of channel selection, the possibility of contention for the same channel by different base stations is avoided, which would be caused by the concurrency of channel assignment among base stations. Interfering base stations do not engage in channel assignment at the same time.
While the above-described scheme satisfactorily addresses the issue of measurement delay and contention, it does not always lead to interference-free channel assignments. Interference can result because the path traversed by the signal from the serving base station to the mobiles served by neighbor base stations, is not the same as the path traversed by the signal transmitted by a neighbor base station and sensed by the mobile during the interference measurement. As illustrated in FIGS. 1 and 2, there could be instances whereby mobile MS1 101 or 201 which is served by base station BS1, does not receive the pilot tones from a neighboring base station, base station BS2, even though the pilot is transmitted. This could be due to an obstruction, or simply to the orientation of a directional antenna. Hence, if base station BS1 selects the channel corresponding to the missed pilot tone, it could cause interference to mobile MS2 103 or 203, which is served by base station BS2.
Thus, a need arises for an interference-sensing scheme for use in asymmetric channel assignment which provides improved reliability and performance over conventional schemes.